Broadcom CEO Responds to Concerns Over AI Slowdown

Hock Tan projects that Broadcom's AI semiconductor revenue will reach $115 billion in fiscal 2027, and further double to $230 billion in fiscal 2028.

Local time on Monday (September 14), Broadcom CEO Hock Tan downplayed market concerns that the development of cutting-edge AI models may slow down. He said in an interview with the media that even if the R&D pace of AI models changes, Broadcom will keep its long-term revenue targets unchanged.

Previously, Dario Amodei, CEO of Anthropic, published an article over the weekend calling for a moderate slowdown in the development speed of AI models.

Amodei's proposition was supported by Sam Altman, CEO of OpenAI, and Elon Musk, the world's richest person, which caused unease among investors and led the market to re-evaluate the future demand for computing power.

Broadcom's share price fell 4.8% on Monday. Anthropic is one of the important customers of Broadcom's custom chip business, so the market is particularly sensitive to Amodei's statement.

Broadcom Sticks to Its AI Revenue Targets

In Monday's interview, when asked whether the discussion about AI slowdown made him reconsider Broadcom's AI semiconductor revenue forecasts for fiscal 2027 and 2028, Hock Tan replied: "Not at all."

Hock Tan said: "We see that the demand for computing power infrastructure will continue to remain strong, whether it is for AI development and cutting-edge model training, or the inference demand after these models are put into practical applications, and I think this demand is highly sustainable."

Hock Tan previously predicted at the earnings call earlier this month that the company's AI semiconductor revenue would reach $115 billion in fiscal 2027, and double again to $230 billion in fiscal 2028. The target for fiscal 2028 is higher than market expectations, which is also one of the main highlights in Broadcom's earnings report for that quarter.

Broadcom's AI revenue mainly comes from two parts: custom AI accelerators, and network chips used in AI systems.

Hock Tan revealed that Anthropic is expected to become Broadcom's largest custom chip customer in 2027 and remain in that position in 2028.

In the past, Google has long been regarded as Broadcom's largest custom chip customer, and the two parties jointly designed Google's Tensor Processing Unit (TPU).

Optimistic About the Sustained Growth of Inference Demand

Hock Tan is particularly optimistic about AI inference demand. He said: "I'm not sure about the training side, but if you want to truly productize inference, I think the demand for this part will still be very, very strong."

Hock Tan also said that he agrees with Amodei's view that AI needs certain restrictions, but he is not so worried about the future development direction of the technology.

"Like any tool, we all need to establish governance mechanisms and security safeguards for how we use it," he said.

But Hock Tan also emphasized: "AI is not a living thing that will get out of control and run around on its own."

Compared with the risks, Hock Tan puts more emphasis on the potential of AI to improve productivity. He said: "AI, generative AI and the creation of these cutting-edge models... will generate huge value." He compared the development of AI to the Industrial Revolution that started in the UK in the 18th century.

"In the final analysis, it is still a tool that will help our society and all mankind reach a higher standard of living," Hock Tan said.

This article is from the WeChat official account "Sci-Tech Innovation Board Daily", Author: Xia Junxiong, published with authorization from 36Kr.